|
What is org.weiss.galapagos?
Galapagos is support code for doing Genetic Algorithms in Java.
I’m implementing this support code with the intended audience
of an undergraduate Artificial Intelligence class (see below).
Galapagos is not just the final project to cap off my CS major.
No, it’s the final product of three years worth of ideas
marinating in the deep recesses of my psyche. To see this project
come to fruition is not only exciting, but (somewhat) cathartic.
Furthermore, I’m very excited to take my old, abstract
ideas and implement them with new technologies. Marrying my
own familiar ideas with my new techniques for a (hopefully)
usable project is one of the many goals of this endeavor.
Galapagos
makes Genetic Algorithms easier
The beauty of Genetic Algorithms is that the searching
process is completely independent of the problem
itself. This dichotomy, then, lends itself easily
to a programming framework. What I will be doing
this year is actually writing the code for Genetic
Algorithms once and for all so that anyone can simply
choose which pieces they want to use.
The notion of “choosing pieces” brings
me to another point about Genetic Algorithms: code
integration. There are many different ways to write
a Genetic Algorithm because there are many ways
to implement the operators. So not only will it
be beneficial to write the “textbook”
operators for the algorithms, the programmer (my
user) will be able to implement their own operator
and use it in conjunction with the rest of the framework. |
Galapagos
is a pedagogical device.
What good is a product without intentionally designating
an audience? Even though Genetic Algorithms programming
libraries have been implemented many times (and
with much more software prowess than that held by
yours truly), I found that most products were not
helpful for students. If a student taking an AI
class wishes to put together a genetic algorithm
in a single afternoon, simply for exploration or
possibly a lab, then using one of those massive
libraries is both too daunting and unhelpful. A
goal, then, for this project is to make Galapagos
programmatically and educationally accessible for
undergraduates. In fact, let’s make it formal:
Goal:
Galapagos shall be able to be used by an
undergraduate student, with a competency in Java,
in a single afternoon.
|
|
Project Components
Here, I will outline some of the (future) technical achievements
that this project will bring. The order below is the order in
which I plan to attack this project:
galapagos.tests
As I develop pet problems for this framework, I will be
creating unit tests to keep everything in shape. Using JUnit,
I actually tend to follow the model that [link]Professor
Frens advocates called “test driven development”
where you write a test first, then write your code to pass
that test. Since Galapagos is a conglomeration of many different
pieces that must work together seamlessly, it makes perfect
test to emphasize unit testing throughout the application.
If applicable, I hope to use mock objects from JMock to
enrich the unit testing. |
galapagos.framework
This is the heart of the project: a programming framework
that allows the user to simply implement their fitness
function and a genome manufacturer and the rest of Genetic
Algorithms will be at their fingertips.
If time allows, I would also like to create support for
distributing the workload of a Genetic Algorithm over
a network. This would involve thread pooling and networking,
but would be very slick. |
galapagos.examples
How far I get on this project is greatly dependent on
how much I end up emphasizing applying Genetic Algorithms
versus developing Galapagos. If I don’t get to the
runner and the reporting, it’s because I found some
considerably stimulating problems along the way. But,
let’s hope I am not so selfish as to not make all
of my work accessible to future undergraduates out there.
Regardless of the emphasis, I plan on writing tutorials
for using the framework. These tutorials will not only
demonstrate my commitment to my future users, they will
also serve as a challenge in educating – a field
in which I hope to find myself someday.
|
galapagos.runner
If a student wishes to simply run a GA without writing any
Java code (i.e. use all of the operators from galapagos.framework),
then all he or she would need to do is specify an XML file
with all of the appropriate properties. But allowing for
XML documents is not enough, I hope to create a simple GUI
to create an XML file. This technology will involve using
the SAX parser along with JavaBeans. |
galapagos.report
The reporting portion has two main mechanisms: one for grabbing
the information from these algorithms and the other is reporting
it. The first mechanism will be incorporated into the framework
in a way that captures information on each genome for each
generation. The output formats of the second mechanism will
be PDF, HTML, and XML. At this amorphous, preliminary stage
of planning, I think I will use JFreeChart and JasperReports
as tools. |
Eclipse
Though not formalized, it is assumed in this project that
the users will be using Eclipse for their Java development
(besides, is there any other way to develop Java anyway??).
Clearly this is not a requirement, but most places use
Eclipse anyway and I’ll be using Eclipse for my
development.
|
Galapagos is Open Source
This project is on SourceForge under the GNU GPL License.
This page was last updated on
June 11, 2006
|
]